Biostatistics

Introduction

  • Pattern → Hypothesis → Evidence → Action


What is Biostatistics?

  • Application of statistical principles to medical, public health, and biological applications

    • THEORY

  • Why does it matter…

    • Evidence-based decisions in medicine, health, nursing, public health, policy


To figure out if a theory is true or not, they will need to create a sample

A sample:

  • Should make inferences about the population


The most popular one is clinical trails (Testing of a vaccine works within a certain population)

  • And when it need to be tested, it will always be on mice


This issues and theories must reach these requirements:

  • Must clearly define a research question

  • Choose an appropriate study design

  • Select a sufficiently large and representative sample

  • Carefully collect and summarize data

  • Must quantify uncertainity/probablity

  • Account for relationships among characteristics

  • Limit references to appropriate population


Biostatistics in action:

  • H1N1 outbreak (Swan flue) → Disease surveillance, tracking spread

  • High-risk health behaviors → Survey data, self-reported samples

  • Risk factors for autism → Pattern detection, confounding variables


Contract tracing

  • Sampling who was exposed to estimate the spread in the population


Main issues of each population:

  • Children

    • Obesity

    • Immunizations

    • Asthma

    • Autism

  • Adolescents

    • Alcohol and tobacco use

    • Depression

    • STDs

    • Traffic accidents

  • Adults

    • Cancer

    • CVD (Cardiovascular disease)

    • Substance abuse

    • HIV/AIDs

    • Mental health


The number one killer of men and women in the United States - CVD


Study Designs


The ratio of Biostatistics:

  • 80% Theory/20% Math


What are study designs?

  • Blueprints for a research/study

    • What? → What are we trying to find out?

    • Who? → Who will we study?

    • How? → How will we collect the data?

    • When? → When will we collect it?

    • Why? → What type of question can the design answer?


Observational Studies: Descriptive

  • Goal: Describe the health status of a population, do not assign an exposure or treatment

  • Purpose: Summarize the data

    • There is no cause-and-effect

  • Data: Surveys, health records, census

  • Pros: Simple, inexpensive

  • Cons: Cannot explain cause


Observational Studies: Analytical

  • Goal: Explore the associations between exposures and outcomes

  • Purpose: Different for each kind of study

  • Two or more variables are related

  • “Is X associated with Y?”

Ex…

    Researchers examine whether college students who sleep less than 6 hours per night are more likely to report anxiety.

Exposure: Sleep?

Outcome?


Case-control Study:

  • Goal: To compare people with disease (cases) vs without (control)

    • Start with a disease and then look back for exposure

  • Purpose: To look backward in time for risk factors

  • Pros: Quick and cheap

  • Cons: Recall bias, cannot prove cause, needs careful selection of cases and controls


Cohort (Time) Study

  • Goal: Follow a group of healthy people over time that meet the inclusion criteria

  • Purpose: Compare exposed vs non-exposed, assess exposure, and evaluate outcome status

    • Start with exposure and follow over time

  • Pros: Strong evidence, can study many outcomes

  • Cons: Expensive, and time-consuming


Cross-sectional Study

  • Goal: To create a snapshot on time

    • Exposure and outcome measured at the same time

  • Purpose: To measure and analyze data from a population at a single, specific point in time

  • Pros: Quick and easy

  • Cons: Cannot show cause and effect, no temporal information, non-response


Experimental Studies

  • Researchers INTERVENE → then observe what happens

    • Gazebo or Placebo effect

  • Researchers:

    • Assign exposures

  • Patients:

    • Divided into two different groups (Treatment or Control)

  • Goal: Determine whether an intervention causes a change in an outcome

  • Strongest evidence for cause-and-effect


RCT - Randomized Controlled Trail

  • GOLD STANDARD

  • Goal: To randomly assign participants too treatment vs control

  • Purpose: To establish a clear cause-and-effect relationship

  • Pros: Strongest evidence for cause

  • Cons: Expensive and ethical limits


Quasi-experimental studies

  • Purpose: An intervention without randomization

  • NOT fully randomized, but still very useful

  • Pros: Easier to implement, and still shows impact

  • Cons: More bias risk


How to decide which design is the best?

  • Depends on the question

  • What is currently known about the topic

  • How common

  • How long will the study take

  • Ethical issues


Examples:

  • Researchers tracked 5,000 adults for 20 years to see if high sugar intake leads to diabetes = Cohort

  • Scientists compared people who had Zika virus with those who did not, looking back at mosquito exposure = Case-control

  • A new cancer drug was randomly tested against a placebo in patients before FDA approval = RCT